AI’s Leap in Healthcare and Biotech: Insights from Our Latest Fireside Chat

AI in life sciences isn’t new. In fact, it’s been quietly reshaping the field long before GPT entered the chat, but the last few years have accelerated both the pace and the stakes. In our latest Averin Webinar, I was joined by three leaders pushing the boundaries of AI in healthcare:

• Andrew Hopkins, founder of Exscientia and now CEO of XYME, focused on enzyme design
• Jared Saul, Chief Medical Officer at Amazon Web Services
• Andrey Dobry, CEO of Bioptic, an AI-native pharma platform

Together, we explored the arc of AI’s evolution — from early Bayesian models to agentic systems capable of autonomous decision-making — and what this means for drug development, infrastructure, and the future of scientific discovery.

Machine-designed Molecules

Andrew reflected on the last two decades of applying machine learning to biology, beginning as early as 2006. His journey — from early predictions and molecule design at Exscientia, through its IPO and eventual acquisition by Recursion — offered firsthand insight into how AI moved from curiosity to creative collaborator. Today, his work has shifted from small molecule design to industrial enzymes, with the belief that AI-designed catalysts could transform sustainability at a global scale.

One thread throughout Andrew’s remarks was that creativity itself is no longer the sole domain of humans. AI now contributes novel molecular structures and designs that are patentable, synthesizable, and functional. GenAI, as he noted, is here — and by the end of the decade, ​“all drugs will be designed by AI.”

Turning Data into Decisions

Jared brought the cloud-native perspective from AWS, where he helps clients — including 19 of the top 20 global pharma companies — navigate AI transformation. A former neuroradiologist and entrepreneur, Jared’s vantage point spans startups and large enterprises alike.

His main observation: large language models (LLMs) are already proving valuable in knowledge mining. LLMs can parse literature, integrate proprietary and public datasets, and surface insights much faster than traditional R&D processes allow. But to go from productivity to true impact, Jared emphasized that better data — curated and structured with AI in mind — remains the critical bottleneck.

He also noted that the next frontier will involve models that not only reason but act — potentially even operating lab equipment and executing physical experiments.

The Rise of Agency: Vertical AI

Andrey’s journey from big tech to biotech is emblematic of the AI field’s convergence. Formerly a Google product director (including launching YouTube’s mobile app and sky rocketing to a billion users), he now leads Bioptic, where the goal is to create an autonomous pharma engine. Think of it as an AI-native biopharma company capable of making investment, scientific, and regulatory decisions in a unified loop.

He shared a compelling framework for how machine learning has progressed:
Prediction → Perception → Generation → Reasoning → Agency.

Today, Andrey’s team is focused on building ​“agentic AI” — systems that not only generate outputs, but plan, reason, and write their own code to pursue drug development goals. His view: we don’t just need better algorithms; we need better data generation companies — ones engineered from the ground up to create vast, low-cost biological datasets that can train next-generation foundation models.

What We’re Watching

Each speaker shared technologies they believe are worth tracking closely:

• Andrew pointed to the return of ​“good old-fashioned AI” — logic-based and evolutionary approaches — which, when combined with today’s scale, may unlock new frontiers in creativity and reliability.
• Jared emphasized the need to move past use-case thinking and imagine how workflows themselves will evolve when AI acts as a proactive scientific collaborator.
• Andrey underscored the importance of agentic systems and synthetic data, with reinforcement learning enabling AI to simulate millions of drug development decisions before a single experiment is run.

What Comes Next

AI’s impact in healthcare isn’t speculative. It’s happening now. But it’s also clear that trust, data quality, and collaborative design with domain experts are critical to success. As Andrey noted, decisions in this industry are often siloed between biology, chemistry, regulation, and business. The promise of AI is not just to make each of these more efficient, but to unify them into integrated decision-making engines.

At Averin, we believe innovation happens when technical capability meets market reality. That means listening, adapting, and building tools that are not just impressive, but practical and impactful. This conversation was a powerful reminder that we’re only scratching the surface of what’s possible.